AuTexTification Dataset (Full data)

Datasets of the AuTexTification shared task at IberLEF 2023. This task aims to boost research on the detection of text generated automatically by text generation models. Participants must develop models that exploit clues about linguistic form and meaning to distinguish automatically generated text from human text.

This dataset includes the training and test splits with labels for all the subtasks and languages. Additionally, each file includes the domain, the model and the prompt used to generate each sample. The model label mapping for subtask 2 is: {"A": "bloom-1b7", "B": "bloom-3b", "C": "bloom-7b1", "D": "babbage",  "E": "curie", "F": "text-davinci-003"}